Optimizing Spectrum Sensing Parameters for Local and Cooperative Cognitive Radios

Transcription

1 Optimizing Spectrum Sensing Parameters for Local and Cooperative Cognitive Radios Ayman A. EI-Saleh Wireless etworks and Communications Group Universiti Kebangsaan Malaysia Bangi Malaysia and Centrefor Broadband Communications (CBC) Multimedia University Cyberjaya Malaysia Iny Mahamod Ismail and Mohd Alauddin Mohd Ali Department ofelectrical Electronics andsystem Engineering Faculty ofengineering andbuilt Envirnoment Universiti Kebangsaan Malaysia Bangi Malaysia ulan. nlv Abstract - In spectrum-sensing cognitive radio a local radio might fail to reliably and efficiently monitor the available spectrum holes. To combat the channel fading suffered by such a radio a cognitive network of cooperative spectrum sensing users is employed. The spectrum sensing time duration should be minimized to maximize the capacity of cognitive radio network. In this paper the twin effects of varying the entire frame duration and sensing time duration is studied. The simulation results show that the cognitive radio network capacity can be maximized by an appropriate selection of total frame duration and an optimized corresponding sensing time. Keywords - Capacity cognitive radio optimization spectrum sensing. 1. Introduction Although the frequency allocation charts are very crowded in most countries around the world wide ranges of already licensed spectrum are underutilized according to actual measurements [1]. To satisfy the dramatic increasing demand for radio spectrum cognitive radio [2] is proposed as a key technology to realize the dynamic spectrum allocation. The cognitive radio was first introduced by J. Mitola as an upgraded version ofthe normal software defined radio (SDR) armed with spectrum sensing capability over three degrees of freedom time frequency and space [2][3]. The spectrum sensing is normally considered as a pure detection problem where the CR-assisted users have to scan a vast range of frequencies to observe available 'white spaces' or 'holes' that are temporarily and spatially available for transmission. The CR-assisted users are classified as secondary users (SUs) competing with primary users (PUs) who are obviously Licensees or alternatively users of existing technologies on unlicensed bands (e.g. IEEE802.11a) [4]. The SUs are allowed to utilize the frequency bands of the PUs when they are not currently being used but they should willingly and quickly vacate the band once a PU has been detected. This fast vacation is necessary to avoid causing harmful interference to the PUs who should maintain ubiquitous and uninterrupted accessibility. Therefore the SUs are required to periodically monitor the PUs activities using fast and reliable detection/sensing algorithms. In such algorithms the PU(s) detection probability and false alarm probability measured by the SU will be ofinterest. In this paper a cognitive radio network based on WRA system deployment has been employed. Each frame in WRA consists of one sensing time slot plus one data transmission slot. The effect of varying the total frame duration and the corresponding sensing time on the SU(s) capacity for local and cooperative spectrum sensing was studied in detail. In local sensing a SU makes an individual decision on the presence of PUs whereas in cooperative sensing several SUs collaborate together to come out with a final decision on the presence of PUs by combining all individual decisions of local SUs at a central base station (BS) using OR or AD fusion schemes [5][6]. The collaborative sensing aims to improve the detection sensitivity at low SR environments as well as to tackle the hidden terminal problem where the PUs activities might be shadowed from the local SU receiver by any existing intermediate objects such as in fading environments [7]. In this work the capacity of SU(s) of local and cooperative sensing is analyzed under two operational modes namely the constant primary user protection (CPUP) and constant secondary user spectrum usability (CSUSU) scenarios. In CPUP scenario the interference from SUs to PUs will be set to a specific level that is low enough to ensure ubiquitous and uninterrupted service for the active PUs. This is done by fixing the probability ofdetecting PUs to a high value e.g. P d = 0.9 while minimizing the probability of false alarm. On the other hand in CSUSU scenario the usability of unoccupied bands by SUs can be kept constant by setting the probability of false alarm at a certain level e.g. PI = 0.1 while maximizing the probability of detection. In this paper the capacity of the SU network is analyzed under these two operational modes. This paper is structured as follows Section 2 reviews the channel sensing hypotheses energy detector and the CPUP and CSUSU transmission modes for local spectrum sensing. In section 3 the cooperative spectrum sensing is presented using OR and AD fusion schemes. The performance of local and cooperative spectrum sensing is characterized and effect of varying the frame duration and number of cooperative users in section 4 and finally the conclusions are drawn in section 5. ISB Feb ICACT 2009

2 2. Local Spectrum Sensing In local sensing each SU senses the spectrum within its geographical location and makes a decision on the presence of primary user(s) based on its own local sensing measurements. A. Channel Sensing Hypotheses Consider a SU in a cognitive radio system sensing a frequency band Wand a the received demodulated signal is sampled at sampling ratels thenis W. Hence the sampled received signal X[ n] at the SU receiver will have two hypotheses as follows 1fo X[n] =W[n] 14 X[n] =W[n] + S[n] ifpu is absent ifpu is present (1) where n = 1... K; K is the number of samples. The noise W[n] is assumed to be additive white Gaussian (AWG) with zero mean and variance a. S[n] is the primary user's signal and is assumed to be a random Gaussian process with zero mean and variancea. The goal ofthe local spectrum sensing is to reliably decide on the two hypotheses with high probability of detection Pd and low probability of false alarm Pf P d and Pf can now be defined as the probabilities that the sensing SU algorithm detects a PU under 1fo and 14 respectively. B. Statistical Model ofenergy Detector The energy detector is known as a suboptimal detector which can be applied to detect unknown signals as it does not require a prior knowledge on the transmitted waveform as the optimal detector (matched filter) does. The decision statistic T for energy detector is given by K T = L (X[n])2 n=l It is well known that under the common eyman-pearson detection performance criteria the likelihood ratio yields the optimal decision. Hence the energy detector performance can be characterized by a resulting pair of (PI' P d) that is estimated as Pf = P(T> P 1110) P d = P(T> P11lj) where P is a particular threshold that tests the decision statistic. Since we are interested in low signal-to-noise ratio of primary user (SR =a 2/a 2) regime large number of p x w samples should be used. Thus the test statistic chi-square (2) (3) distribution can be approximated as Gaussian based on the central limit theorem. Then C. Cognitive Radio Transmission Scenarios 1) Constant Primary User Protection (CPUP) Scenario This transmission mode is viewed from the PUs' perspective. It guarantees a minimum level of interference to PUs who by right should not be affected by the SUs transmission. This scenario can be realized by fixing Pd at a satisfactory level e.g. 90% and trying to minimize Pf as much as possible. Thus PI is derived to be where the number of samples K is the product of sensing time times sampling frequency. 2) Constant Secondary User Spectrum Utilization (CSUSU) Scenario This mode is taken from the SUs' perspective; it aims to standardize the spectrum utilization by SUs. As such the P f values should be fixed at lower values (e.g. 10%) while keep maximizing P d which can be written in terms of a desired Pfas follows Q-l{P f )-lsr p (l+sr p ) 3. Cooperative Spectrum Sensing The collaborative sensing aims to improve the detection sensitivity at low SR environments as well as to tackle the hidden terminal problem where the PUs activities might be shadowed from the local SU receiver by any existing intermediate obstacles. This section presents the SU cognitive radio network model using some well-known fusion schemes. In addition the overall network PU detection and false alarm probabilities will be derived for the CPUP and CSUSU transmission scenarios respectively. (4) (5) (7) ISB Feb ICACT 2009

3 A. Cognitive Radio etwork Deployment The network deployment in this paper is based on the IEEE WRA [8]. The WRA base BS collects information on the PU activities from the SUs within its coverage area as shown in Figure 1. Local SUs keep monitoring the presence of a PU which is a TV broadcast station and send their detection and false alarm probabilities to the base station for combining them into one overall final decision. In this scenario it is assumed that the TV BS is far away from the WRA BS and therefore low SRp values are used.. yeq ft...UC;O I" J " " J " J... "... " I --- \ B TV BS=PU WRA users = SUs P f = TIP f ; ;=1 C. Derivations ofetwork Probabilities under CPUP and CSUSU Scenarios In this section the SUs network false alarm and detection probability formulas will be stated under CPUP and CSUSU scenarios respectively. Let's here for instance assume that we are interested to derive the overall false alarm probability of the network PI under the CPUP transmission mode using OR fusion scheme. Firstly we find the individual desired detection probability P d j in terms of the desired network detection probability P d using (8). Secondly the probability of false alarm of each SU Pfi can be found by substituting the P di equation into (6). Finally PI is estimated by substituting the Pfi for CPUP-OR combination is (11 ) equation into (9). Thus P.l Figure 1. A Simplified Representation of an IEEE WRA System Deployment. B. Fusion Schemes for Local Spectrum Sensor' Decisions At the SUs base station all local sensing information are combined and merged into one final decision using Chair Varshney fusion schemes [5][ 6]. Two fusion schemes are used in this paper OR- and AD-rule. In OR-rule fusion scheme the final decision on the presence of a PU will be positive if only one SU of all collaborating users detects this PU. Assuming that all decisions are independent the detection and false alarm probability of the SUs network under OR-rule Pd and P.l respectively can then be mathematically written as Pd = 1- IT (1 - Pdi ) i=1 P f = 1-n(l- PfJ i=1 where P di and Pfi are the individual detection probability and false alarm probability respectively. is the number of cooperating SUs. In AD-rule fusion scheme all collaborating SUs should declare the presence of a PU in order for the final decision to be positive. Again assuming that all decisions are independent the SUs network probabilities under AD-rule can be presented as (8) (9) Please refer to [9] for more explanations on the overall network probabilities of the other three combinations CPUP_AD CSUSU-OR and CSUSU-AD. Thus we introduce PI for CPUP-AD combination can be derived as In CSUSU scenario the false alarm probability of the SUs network is set constant atpf' and the detection probability of the SUs network Plh is derived. For CSUSU-OR ==I-TI 1- i Similarly for CSUSU-AD Q-l(l--Pf))-SRpj (I+SJ (14) Pd =Ilpdi i=l (10) ISB Feb CACT 2009

4 4. Optimizing Sensing Parameters for Local and Cooperative Spectrum Sensing In this section we analyze the relationship between SUs capacity and sensing capability for both local and cooperative sensing under the CPUP and CSUSU transmission modes. In WRA system each frame consists of one sensing slot (ts) plus one data transmission slot (1j - t s) where 1j is the total frame duration. Indeed short sensing slots should be always targeted as it results in longer data transmission slot and therefore higher throughput capacity. A. Problem Formulation There are two cases for which the SUs network might operate at the PU's licensed band; the first one is when the PU is inactive and the SUs successfully declare that there is no PU. In this case the normalized capacity of the WRA system is represented as Co =[1- }-Pf)P(JlO) (16) where P(Jfo) is the probability that the PU is inactive in the frequency band being sensed. The other case is when the PU is active but the SUs fail to detect it. The normalized capacity is then given by C1 =[1- ij(1- P d )P(Jij) (17) where P(Hi) is the probability of the PU being active in the frequency band of interest. Obviously P(Jfo) + P(%.) = 1. The objective of this research is to determine the optimal sensing time for each frame such that the SUs network capacity is maximized. Consequently this objective can be formed as an optimization problem described as follows Subject to 0 < Is S T I and P d Pd under CPUP or PI S Plunder CSUSU B. Optimizing Sensing Parameters for Local Spectrum Sensing (18) In this section MATLAB simulations have been performed to investigate the capacity-sensing capability optimization problem. The WRA frame duration was set to 100 ms and the one- side bandwidth of PU bandpass signal is selected to be 3MHz. The S is set to -18 db. Let's first define the optimal sensing time by the sensing time at which the SU capacity is maximized. For local spectrum sensing under CPUP transmission scenario (6) states that Pf decreases with increasing the sensing time and consequently Pf suppose to increase the SU capacity. However Figure 2 shows that decreasing Pfdoes not lead to an absolute increase in the SU capacity as thought but instead there is an optimal sensing time at which the capacity is maximized. Table 1 summarizes the maximum capacities obtained in Figure 2 and their corresponding optimum sensing times at different detection probabilities and total frame durations. It is seen that for a given total frame time duration let say 100 ms increasing the detection probability from 0.7 to 0.9 leads to degrade the capacity in general and the maximum normalized capacity decreases from to The optimum sensing time is then increased from 2.85 ms to 4.5 ms. On the other hand if the detection probability has been set at a certain value let say 0.9 for high PU protection then increasing the total frame time duration from 100 ms to 500 ms leads to increase the optimum capacity from to and also increase the optimum sensing time from 4.5 ms to a range of 5.8 to 6.1 ms. Obviously increasing the total frame duration leads to a very interesting finding that is the SU capacity becomes more uniform r-----r ;;.;> ;.;.;;.;.; f Jo; vitl;_ -T- "'0 i i 0 I... I rttj-> o i Z i Pd = 0.7 &Tf = 100 ms '; i _._._.. Pd = 0.9 & Tf= 100 ms Pd = 0.7 &Tf = 500 ms i Pd =0.9 & Tf=500 ms Sensing time (ms) Figure. 2 The Effect of Varying the Detection Probability and Total Frame Duration on the ormalized Capacity versus Sensing Time Relation under CPUP Transmission Mode. Under CSUSU scenario (7) reveals that Pd increases with increasing the sensing time this means that the PU will be more protected but unfortunately the SU capacity will then decrease as shown in Figure 3. It is shows that at certain frame duration let say 100 ms decreasing the false alarm probability from 0.3 to 0.1 leads to increase the SU capacity in general. This matches the fact that there will be more chances for SUs to use the spectrum if the probability to detect a PU while it is absent is low. On the other hand ifthe false alarm probability has been set at a specific value say 0.1 for high spectrum usability by SUs then increasing the 20 ISB Feb CACT 2009

5 total frame duration increases the capacity and makes it more uniform. This finding matches the one above for the CPUP transmission scenario. Table 1. Performance Evaluation for Local Spectrum Sensing under CPUP Transmission Mode Detection Probability Pd Frame Duration Tf (ms) Optimum Sensing Time t s (ms) Maximum Capacity C (bits/sec/hz) i - J - -;I t ' ; '..."..".".L " r " ".o.-.o' !---u I ". f =;I=;J.---- E Pf=0.1&Tf=100ms 0.4 sensing time respectively. Figure 4 shows that the maximum SUs network capacity increases by cooperating more users in the network using OR and AD fusion schemes. In addition it is also obvious that the maximum normalized improves by lengthening the total frame duration from 100 ms to 500 ms. Interestingly at long frame duration the network capacity is maintained at a constant level and it is going to be significantly changed by incorporating more users. Figure 5 reveals that cooperating more users will reduce the optimal sensing time required to achieve the maximum throughput. It is also seen that lengthening the frame duration will logically increase the optimal sensing time. j l i l L f - - OR-rule & Tf= 100 ms \ AD-rule & Tf 100 ms E 0.78! OR-rule & Tf= 500 ms AD-rule & Tf = 500 ms l5--===1cO======I15======J20 20 Sensing time (ms) Figure 3 The Effect of Varying the False Alarm Probability and Total Frame Duration on The ormalized Capacity Versus Sensing Time Relation under CSUSU Transmission Mode. C. Optimizing Sensing Parameters for Cooperative Spectrum Sensing The cooperative users are actually local sensors cooperate together within the coverage area of a BS for reliable PU detection. They are to perform individual repetitive sensing cycles and send their individual decisions to the WRA BS as individual detection and false alarm probabilities. The sensing time period which is a fraction of total frame time transmitted by the SU network should be always as minimal as possible to maximize the SU network capacity. In order to estimate the capacity of WRA network under let say CPUP scenario we should first determine the overall PI of the network using (12) or (13) for OR or AD fusion schemes respectively. Then the estimated P.f together with the desired fixed Pd are substituted in (18) to calculate the overall capacity of the network. Similar procedure applies for CSUSU scenario. In this section the number of cooperating SUs is varied from 1 user (no cooperation) to 20 users (all available users in the network are cooperating). The optimal sensing time here is again defined as the sensing time duration at which the SUs network capacity is maximized. Let us first consider the CPUP transmission mode. Figures 4 and 5 presents the effect of varying the number of cooperative users as well as the total frame duration on the maximum capacity obtained and its corresponding optimal ISB E i---._--j I < /.-- ; Pf = 0.3 & Tf = 100 ms --_ --- _ f Pf= 0.1 & Tf= 500 ms Pf = 0.3 & Tf = 500 ms i 0.3 L======I========r= o 5 10 I = umber of users. Figure 4 Maximum ormalized Capacity versus umber of Users under CPUP Mode Using Logical OR and OR Fusion Schemes. 7 r---;----;=!=========!=========;- i - OR-rule & Tf= 100 ms t AD-rule & Tf = 100 ms OR-rule & Tf= 500 ms \ AD-rule & Tf 500 ms E \ 5 = umber of users. 20 Figure 5 Optimal Sensing Time versus umber of Users under CPUP Mode Using Logical OR And OR Fusion Schemes. Figures 6 and 7 present the CSUSU transmission mode using either OR or AD fusion respectively. For both cases It was found that at short sensing times e.g. ts is less than 5 ms of total frame duration cooperating more users will significantly reduce the network capacity whereas at longer sensing times there was no effect on the network capacity by incorporating more users in the cognitive network. On the Feb CACT 2009

6 other hand as the total frame duration increases the network capacity is improved and again the network capacity variation becomes lesser with increasing the sensing time f=500 ms --Tf= 100 ms 0.64 '-----'-2---4"" l6' '10 Sensing time ofthe ith user (ms). where i =1 Figure 6 ormalized Capacity versus Sensing Time for Users under CSUSU Mode Using Logical AD Fusion Scheme (The curves at = 3 to 19 are not indexed to improve the visibility) ff=500 ms --Tf= 100 ms 0.64 L...---L I.-4---S...l S..L I10 Sensing time ofthe ith user (ms) where i =1 Figure 7 ormalized Capacity versus Sensing Time for Users under CSUSU Mode Using Logical OR Fusion Scheme (The curves at = 3 to 19 are not indexed to improve the visibility). 5. Conclusion This research work presents a performance evaluation for local and cooperative spectrum sensing in cognitive radio networks under CPUP and CSUSU transmission modes by varying the sensing time total frame duration and number of cooperative users. In CPUP mode it was found that there is an optimal sensing time at which the capacity of local or cooperative users is maximized. Increasing the number of cooperative users lead to increase the maximum capacity of cooperative SUs network and decrease the corresponding optimal sensing time. In CSUSU mode there was no such an optimal sensing time as in CPUP and the capacity of local or cooperative users keeps decreasing as the sensing time increases. In addition it was observed that ifthe sensing time is Iess than 5 ms of the total frame duration incorporating more users in the network leads to decrease its capacity whereas at longer sensing times incorporating more users has no effect on the network capacity. For both CPUP and CSUSU modes employing long frame duration will always improve and stabilize the network capacity and the demand for incorporating more users becomes less critical. This is evidently a tradeoffcase between the total frame duration and number ofcooperative users in the network. Acknowledgment This research work done in this paper was sponsored by UKM-OUP-BT /2008 research fund The authors would like to thank and appreciate Professor Kasmiran Jumari for his kindness valuable assistance and financial support. References [1] Federal Communications Commission "Spectrum policy task force report FCC " ov [2] 1. Mitola and G. Q. Maguire "Cognitive Radios making software radios more personal" IEEE personal communications vol. 6 no. 4 pp Aug [3] 1. Mitola "Cognitive radio an integrated agent architecture for software defined radio" PhD thesis KTH Royal Institute of Technology Stockholm Sweden [4] A. A. EI-Saleh M. Ismail o. B. A. Ghafoor and A. H. Ibrahim "Comparison between Overlay Cognitive Radio and Underlay Cognitive Ultra Wideband Radio for Wireless Communications" Proc. of the Fifth lasted (AsiaCS 2008) pp April Langkawi Malaysia. [5] Z. Chair and P. K. Varshney "Optimal data fusion in multiple sensor detection systems" IEEE Trans. on Aerospace and Elect. Syst. vol.22 pp January [6] P. K. Varshney "Distributed Detection and Data Fusion". Springer [7] A. Ghasemi and E.S. Sousa Collaborative spectrum sensing for opportunistic access in fading environments Proc. of DySPA'05 ovember [8] IEEE wireless RA "Functional requirements for the WRA standard IEEE /0007r46" Oct [9] A. A. EI-Saleh M. Ismail M. A. M. Ali and A.. H. Alnuaimy "Capacity Optimization for Local and Cooperative Spectrum Sensing in Cognitive Radio etworks" Proc. of ICCT February ISB Feb CACT 2009